Lung Area Research Articles

POLCOVID: a multicenter multiclass chest X-ray database (Poland, 2020–2021)

The outbreak of the SARS-CoV-2 pandemic has put healthcare systems worldwide to their limits, resulting in increased waiting time for diagnosis and required medical assistance. With chest radiographs (CXR) being one of the most common COVID-19 diagnosis methods, many artificial intelligence tools for image-based COVID-19 detection have been developed, often trained on a small number of images from COVID-19-positive patients. Thus, the need for high-quality and well-annotated CXR image databases increased. This paper introduces POLCOVID dataset, containing chest X-ray (CXR) images of patients with COVID-19 or other-type pneumonia, and healthy individuals gathered from 15 Polish hospitals. The original radiographs are accompanied by the preprocessed images limited to the lung area and the corresponding lung masks obtained with the segmentation model. Moreover, the manually created lung masks are provided for a part of POLCOVID dataset and the other four publicly available CXR image collections. POLCOVID dataset can help in pneumonia or COVID-19 diagnosis, while the set of matched images and lung masks may serve for the development of lung segmentation solutions.

PULM4: Silicone-Based Hollow Fiber Membrane Lungs For Long-Term Extracorporeal Respiratory Supports: Initial Results From A Pilot Study Using A Goat Model

Background: There has been an increasing demand of organs for lung transplant to save those who suffer from end-stage lung disease. However, the number of organ donations have never matched that of candidate recipients, leading to a high mortality rate on the waiting period for transplant. One potential approach to decrease the wait-list mortality is to develop an artificial gas-exchange device that can support lung function until the time of transplantation. We’ve started to develop a gas-exchange system that consists of silicone-based membrane lungs with methacryloyloxyethyl phosphorylcholine (MPC) polymer coating and tubing system for a longer-term use. Objective: The purpose of this study was to evaluate the long-term durability, oxygenation, and pressure loss of newly developed artificial lungs using a large animal model. Materials and methods: The surface area of the membrane lung is approximately 2 m2. Central ECMO from the pulmonary artery to the left atrium was introduced to a healthy female goat via left thoracotomy under general anesthesia (N=3). Unfractionated heparin infusion was utilized for anticoagulation (target activating clotting time: 150-200 sec). After recovery from the surgery, blood gas and flow pressure before and after the artificial lung with flow rates of 1, 2, and 3 L/min were tested weekly. Daily physiological data, regular blood biochemistry and coagulation tests, and histopathological findings after autopsy were also evaluated. Thrombus adhesion to the silicone membrane was evaluated using an electron microscopy after the experiments. Results: Durations for testing the artificial lungs were 53, 58, and 100 days. The experiments were terminated due to an accidental tubing event, a sudden cardiac complication, and massive renal hemorrhage. Averages of oxygen transfer rates (ml/min) and oxygen pressure gradients (mmHg) at each flow rate were 63.3 and 466.5 at 1 L/min, 118.3 and 291.3 at 2 L/min, and 152.7 and 138.2 at 3 L/min. Averages of flow pressures through the artificial lungs at each flow rate were 21.7 mmHg at 1 L/min, 44.7 at 2 L/min, and 66.9 at 3 L/min. No significant increase of pressure loss or decrease of oxygenation was observed throughout the experimental periods. Thrombus at the inlet of the membrane was mainly observed, but those inside and near the outlet were minor, indicating the long-term antithrombotic property of the polymer coating. Electron microscopical examinations revealed minimal thrombus formation on the hollow fiber membrane inside the lung. At autopsy, although micro-infarction and thrombus were observed in the heart and kidney, native lungs were well maintained without appreciable congestion. Conclusion: The newly developed MPC polymer-coated silicone-based lungs could be used stably for more than 50 days. Improving the oxygenation in higher blood flow rates with reduced pressure loss is our ongoing subject to work on.

Comparison of artificial intelligence versus real-time physician assessment of pulmonary edema with lung ultrasound

BackgroundLung ultrasound can evaluate for pulmonary edema, but data suggest moderate inter-rater reliability among users. Artificial intelligence (AI) has been proposed as a model to increase the accuracy of B line interpretation. Early data suggest a benefit among more novice users, but data are limited among average residency-trained physicians. The objective of this study was to compare the accuracy of AI versus real-time physician assessment for B lines. MethodsThis was a prospective, observational study of adult Emergency Department patients presenting with suspected pulmonary edema. We excluded patients with active COVID-19 or interstitial lung disease. A physician performed thoracic ultrasound using the 12-zone technique. The physician recorded a video clip in each zone and provided an interpretation of positive (≥3 B lines or a wide, dense B line) or negative (<3 B lines and the absence of a wide, dense B line) for pulmonary edema based upon the real-time assessment. A research assistant then utilized the AI program to analyze the same saved clip to determine if it was positive versus negative for pulmonary edema. The physician sonographer was blinded to this assessment. The video clips were then reviewed independently by two expert physician sonographers (ultrasound leaders with >10,000 prior ultrasound image reviews) who were blinded to the AI and initial determinations. The experts reviewed all discordant values and reached consensus on whether the field (i.e., the area of lung between two adjacent ribs) was positive or negative using the same criteria as defined above, which served as the gold standard. Results71 patients were included in the study (56.3% female; mean BMI: 33.4 [95% CI 30.6–36.2]), with 88.3% (752/852) of lung fields being of adequate quality for assessment. Overall, 36.1% of lung fields were positive for pulmonary edema. The physician was 96.7% (95% CI 93.8%–98.5%) sensitive and 79.1% (95% CI 75.1%–82.6%) specific. The AI software was 95.6% (95% CI 92.4%–97.7%) sensitive and 64.1% (95% CI 59.8%–68.5%) specific. ConclusionBoth the physician and AI software were highly sensitive, though the physician was more specific. Future research should identify which factors are associated with increased diagnostic accuracy.

Computed Tomographic and Histopathologic Studies of Lung Function Immediately Post Natum in Canine Neonates.

The lung tissue in newborn canine neonates is still in a morphologically and functionally immature, canalicular-saccular stage. True alveoli are only formed postnatally. The aim of this study was to analyze the spatial and temporal development of the ventilation of the lung tissue in vital canine neonates during the first 24 h post natum (p.n.). Forty pups (birth weight Ø 424 g ± 80.1 g) from three litters of large dog breeds (>20 kg live weight) were included in the studies. Thirty-three pups (29 vital, 2 vitally depressed, 2 stillborn neonates) originated from controlled, uncomplicated births (n = 3); moreover, six stillborn pups as well as one prematurely deceased pup were birthed by other dams with delivery complications. Computed tomography (CT) was used in 39 neonates, and histopathologic tissue classification techniques (HALO) were used in 11 neonates (eight stillborn and three neonates died early post natum, respectively) to quantify the degree of aerated neonatal lung tissue. It was shown that, in vital born pups, within the first 10 min p.n., the degree of ventilation reached mean values of -530 (±114) Hounsfield units (HU) in the dorsal and -453.3 (±133) HU in the ventral lung area. This is about 75-80% of the final values obtained after 24 h p.n. for dorsal -648.0 (±89.9) HU and ventral quadrants -624.7 (±76.8) HU. The dorsal lung areas were always significantly better ventilated than the ventral regions (p = 0.0013). CT as well as histopathology are suitable to clearly distinguish the nonventilated lungs of stillborns from neonates that were initially alive after surviving neonatal respiratory distress syndrome but who died prematurely (p = 0.0398). The results of this study are clinically relevant since the lung tissue of canine neonates presents an aeration profile as early as 10 min after birth and continues progressively, with a special regard to the dorsal lung areas. This is the basis for resuscitation measures that should be performed, preferably with the pup in the abdomen-chest position.

Vision Transformers for Lung Segmentation on CXR Images.

Accurate segmentation of the lungs in CXR images is the basis for an automated CXR image analysis system. It helps radiologists in detecting lung areas, subtle signs of disease and improving the diagnosis process for patients. However, precise semantic segmentation of lungs is considered a challenging case due to the presence of the edge rib cage, wide variation of lung shape, and lungs affected by diseases. In this paper, we address the problem of lung segmentation in healthy and unhealthy CXR images. Five models were developed and used in detecting and segmenting lung regions. Two loss functions and three benchmark datasets were employed to evaluate these models. Experimental results showed that the proposed models were able to extract salient global and local features from the input CXR images. The best performing model achieved an F1 score of 97.47%, outperforming recent published models. They proved their ability to separate lung regions from the rib cage and clavicle edges and segment varying lung shape depending on age and gender, as well as challenging cases of lungs affected by anomalies such as tuberculosis and the presence of nodules.

Preoperative assessment of pleural adhesions in patients with lung cancer based on quantitative motion analysis with dynamic chest radiography: A retrospective study.

Preoperative assessment of pleural adhesion is crucial for appropriate surgical planning. This study aimed to quantitatively evaluate the usefulness of motion analysis using dynamic chest radiography (DCR) for assessing pleural adhesions. Sequential chest radiographs of 146 lung cancer patients with or without pleural adhesions (n=25/121) were obtained using a DCR system during respiration (registration number: 1729). The local motion vector was measured, and the percentage of poor motion area to the maximum expiration lung area (%lung area with poor motion) was calculated. Subsequently, percentage values ≥49.0% were considered to indicate pleural adhesions. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated to assess the prediction performance. The percentage of lung area with poor motion was compared between patients with and without pleural adhesions (p<0.05). DCR-based motion analysis correctly predicted pleural adhesions in 21 out of 25 patients, with 47 false-positive results (sensitivity, 84.0%; specificity, 61.2%; PPV, 30.9%; NPV, 94.9%). The lung with pleural adhesions showed a significantly greater %lung area with poor motion than the opposite lung in the same patient, similar to the cancerous lung in patients without pleural adhesions. On DCR-based motion analysis, pleural adhesions could be indicated by an increase in the percentage of lung area with poor motion. Although the proposed method cannot identify the exact location of pleural adhesions, information regarding the presence or absence of pleural adhesions provided by DCR would help surgeons prepare for challenging surgeries and obtain informed consent from patients.

Angiopoietin-like 4 Is a Critical Regulator of Fibroblasts during Pulmonary Fibrosis Development.

Idiopathic pulmonary fibrosis (IPF) is a chronic, progressive, and irreversible interstitial pneumonia caused by the excessive production and deposition of extracellular matrix components, including type I collagen. Activated fibroblasts, called α-SMA (α-smooth muscle actin)-expressing myofibroblasts, are the major source of type I collagen in pulmonary fibrosis (PF), but the mechanisms underlying disease progression have not been fully elucidated. Here, we obtained lung fibroblasts from patients with IPF from both nonfibrotic and fibrotic areas as determined by a lung computed tomography scan and compared gene expression between these areas by DNA microarray. We found that ANGPTL4 (angiopoietin-like 4) was highly expressed only in fibroblasts from the fibrotic area. ANGPTL4 was selectively expressed in the fibroblastic area of IPF lungs, where the myofibroblast marker α-SMA was also expressed. ANGPTL4 also regulates the gene expression of fibrosis-related markers, cell migration, and proliferation. In addition, ANGPTL4 expression in a murine model of PF induced by treatment with bleomycin was significantly induced in the lungs from the acute to the chronic phase. Single-cell transcriptome analysis during the course of bleomycin-induced PF revealed that Angptl4 was predominantly expressed in the activated fibroblasts and myofibroblasts. Moreover, the administration of recombinant ANGPTL4 to the bleomycin-induced fibrosis model significantly increased collagen deposition and exacerbated the PF. In contrast, the pathogenesis of PF in Angptl4-deficient mice was improved. These results indicate that ANGPTL4 is critical for the progression of PF and might be an early diagnostic marker and therapeutic target for IPF.

Depthwise convolution based pyramid ResNet model for accurate detection of COVID-19 from chest X-Ray images

ABSTRACT The global pandemic of coronavirus disease 2019 (COVID-19) causes severe respiratory problems in humans. The Chest X-ray (CXR) imaging technique majorly assists in detecting abnormalities in the chest and lung areas caused by COVID-19. Hence, developing an automatic system for CXR-based COVID-19 detection is vital for disease diagnosis. To accomplish this requirement, an enhanced Residual Network (ResNet) model is proposed in this paper for accurate COVID-19 detection. The proposed model combines the Depthwise Separable Convolutional ResNet and Pyramid dilated module(DSC-ResNet-PDM) for deep feature extraction. Employing the DSC layer minimizes the number of parameters to mitigate the overfitting issue. Further, the pyramid dilated module is used for extracting multi-scale features. The extracted features are finally fed into the optimized Medium Gaussian kernel Support Vector Machine classifier (MGKSVM) for COVID-19 detection. The proposed model attained an accuracy of 99.5%, which is comparatively higher than the standard ResNet50 and ResNet101 models.

Structural priors represented by discrete cosine transform improve EIT functional imaging.

Structural prior information can improve electrical impedance tomography (EIT) reconstruction. In this contribution, we introduce a discrete cosine transformation-based (DCT-based) EIT reconstruction algorithm to demonstrate a way to incorporate the structural prior with the EIT reconstruction process. Structural prior information is obtained from other available imaging methods, e.g., thorax-CT. The DCT-based approach creates a functional EIT image of regional lung ventilation while preserving the introduced structural information. This leads to an easier interpretation in clinical settings while maintaining the advantages of EIT in terms of bedside monitoring during mechanical ventilation. Structural priors introduced in the DCT-based approach are of two categories in terms of different levels of information included: a contour prior only differentiates lung and non-lung region, while a detail prior includes information, such as atelectasis, within the lung area. To demonstrate the increased interpretability of the EIT image through structural prior in the DCT-based approach, the DCT-based reconstructions were compared with reconstructions from a widely applied one-step Gauss-Newton solver with background prior and from the advanced GREIT algorithm. The comparisons were conducted both on simulation data and retrospective patient data. In the simulation, we used two sets of forward models to simulate different lung conditions. A contour prior and a detail prior were derived from simulation ground truth. With these two structural priors, the reconstructions from the DCT-based approach were compared with the reconstructions from both the one-step Gauss-Newton solver and the GREIT. The difference between the reconstructions and the simulation ground truth is calculated by the ℓ2-norm image difference. In retrospective patient data analysis, datasets from six lung disease patients were included. For each patient, a detail prior was derived from the patient's CT, respectively. The detail prior was used for the reconstructions using the DCT-based approach, which was compared with the reconstructions from the GREIT. The reconstructions from the DCT-based approach are more comprehensive and interpretable in terms of preserving the structure specified by the priors, both in simulation and retrospective patient data analysis. In simulation analysis, the ℓ2-norm image difference of the DCT-based approach with a contour prior decreased on average by 34% from GREIT and 49% from the Gauss-Newton solver with background prior; for reconstructions of the DCT-based approach with detail prior, on average the ℓ2-norm image difference is 53% less than GREIT and 63% less than the reconstruction with background prior. In retrospective patient data analysis, the reconstructions from both the DCT-based approach and GREIT can indicate the current patient status, but the DCT-based approach yields more interpretable results. However, it is worth noting that the preserved structure in the DCT-based approach is derived from another imaging method, not from the EIT measurement. If the structural prior is outdated or wrong, the result might be misleadingly interpreted, which induces false clinical conclusions. Further research in terms of evaluating the validity of the structural prior and detecting the outdated prior is necessary.

CT quantitative analysis of pulmonary changes in rheumatoid arthritis.

To explore the application of quantitative computed tomography (CT) in analyses of lung changes in patients with rheumatoid arthritis (RA). A total of 150 clinically diagnosed RA patients underwent chest CT and 150 matched non-smokers subjects with normal chest CT are enrolled. A CT software is applied to analyze CT obtained from both groups. The quantitative indices of emphysema are expressed as the percentage of lung area with attenuation < -950HU to the total lung volume (LAA-950%), and pulmonary fibrosis was expressed as the percentage of lung area with a attenuation of -200 to -700HU to the total lung volume (LAA-200--700%), quantitative indicators of pulmonary vascular include aortic diameter (AD), pulmonary artery diameter (PAD), the ratio of PAD to AD (PAD/AD ratio), the number of blood vessels (TNV), and the cross area of blood vessels (TAV). The receiver operating characteristic (ROC) curve is used to evaluate the ability of these indexes in identifying the changes in the lung in RA patients. Compared to the control group, the RA group has significantly lower TLV, larger AD, and smaller TNV and TAV (3921±1101 vs. 4490±1046, 33.26±4.20 vs. 32.95±3.76, 13.14±4.93 vs. 17.53±3.34, and 96.89±40.62 vs. 163.32±34.97, respectively, with all p < 0.001). Peripheral vascular indicator TAV has the better ability to identify lung changes in RA patients (area under ROC curve AUC = 0.894) than TNV (AUC = 0.780) or LAA-200&sim-700% (AUC = 0.705). Quantitative CT can detect changes in lung density distribution and peripheral vascular injury in patients with RA and assess the severity.

26 An Advanced Triple Cell co-Culture of the Alveolar Epithelial Barrier for Inhalation Hazard Assessment in Vitro

Abstract In vitro lung models are a valuable tool to assess the inhalation hazard of air pollutants, bioaerosols, chemicals and nanomaterials. Though it is essential that these models are characterised and anatomically relevant to the area of the human lung they seek to represent. Previously within inhalation toxicology, focus has been placed on the use of monocultures of alveolar epithelial cells, however use of alveolar co-culture models have been presented as more realistic and sensitive models. Here, we present the development of a novel triple cell co-culture system, cultured at the air-liquid interface that is sensitive to LPS and particulate matter. This co-culture contains alveolar type 1 cells (hAELVi), alveolar type 2 cells (NCI-H441) and alveolar macrophages (differentiated THP-1 (dTHP-1) using PMA (20 nM)). Using CellTrackers™ with confocal microscopy, the ratio of hAELVi cells to NCI-H441 cells was confirmed to be a histologically relevant 16:1 ratio forming a tight monolayer shown through blue dextran membrane permeability assay. Macrophages were seeded on top of the epithelial co-culture at an anatomically relevant density of 1 macrophage per 18x103 μm2 (confirmed via confocal microscopy) to form the triple cell co-culture. The model, when challenged, released more interleukin-8 and TNF-α than the epithelial mono- and bi-culture systems when exposed to LPS (1 μg/mL for 24 hrs) (p&amp;lt;0.05), suggesting that the macrophages do mount the theoretical (pro-)inflammatory response. This anatomically relevant in vitro model forms the basis to investigate inhaled exogenous substances, such as the differential toxicology impacts of particulate matter and nitrogen dioxide.

Feasibility of dynamic chest radiography to calculate lung volumes in adult people with cystic fibrosis: a pilot study

IntroductionDynamic chest radiography (DCR) is a novel, low-dose, real-time digital imaging system where software identifies moving thoracic structures and can automatically calculate lung areas. In an observational, prospective, non-controlled, single-centre...

Ultrasonic evaluation of fetal lung development using deep learning with graph

BackgroundThe neonatal respiratory morbidity that was primarily caused by the immaturity of the fetal lung is an important clinical issue in close relation to the morbidity and mortality of the fetus. In clinics, the amniocentesis has been used to evaluate the fetal lung maturity, which is time-consuming, costly and invasive. As a non-invasive means, ultrasonography has been explored to quantitatively examine the fetal lung in the past decades. However, existing studies required the contour of the fetal lung which was delineated manually. This may lead to significant inter- and intra-observer variations. MethodsWe proposed a deep learning model for automated fetal lung segmentation and measurement, which was constructed combined U-Net with Graph model and pre-trained Vgg-16 network. The graph connection would extract stable feature for final segmentation and pre-trained method could speed up convergence.The model was trained with 3500 datasets augmented from 250 ultrasound images with both the fetal lung and heart delineated manually, and tested on 50 ultrasound images. In addition, the correlation between the size of fetal lung/heart as delineated by the model with gestational age was analyzed. ResultsThe fetal lung and cardiac area were segmented automatically with the accuracy, average Intersection over Union(IoU), sensitivity and precision being 0.991, 0.818, 0.909 and 0.888, respectively. In addition, the size of fetal lung/heart was well correlated with the gestational age, demonstrating good potentials for assessing the fetal development. ConclusionsThis study proposed a new robust method for automatic fetal lung segmentation in ultrasound images using Vgg16-GCN-UNet. Our proposed method could be utilized potentially not only to improve existing research in quantitative analyzing the fetal lung using ultrasound imaging technology, but also to alleviate the labor of the clinicians in routine measurement of the fetal lung/cardiac.

Interstitial Lung Abnormalities at CT in the Korean National Lung Cancer Screening Program: Prevalence and Deep Learning-based Texture Analysis.

Background Interstitial lung abnormalities (ILAs) are associated with worse clinical outcomes, but ILA with lung cancer screening CT has not been quantitatively assessed. Purpose To determine the prevalence of ILA at CT examinations from the Korean National Lung Cancer Screening Program and define an optimal lung area threshold for ILA detection with CT with use of deep learning-based texture analysis. Materials and Methods This retrospective study included participants who underwent chest CT between April 2017 and December 2020 at two medical centers participating in the Korean National Lung Cancer Screening Program. CT findings were classified by three radiologists into three groups: no ILA, equivocal ILA, and ILA (fibrotic and nonfibrotic). Progression was evaluated between baseline and last follow-up CT scan. The extent of ILA was assessed visually and quantitatively with use of deep learning-based texture analysis. The Youden index was used to determine an optimal cutoff value for detecting ILA with use of texture analysis. Demographics and ILA subcategories were compared between participants with progressive and nonprogressive ILA. Results A total of 3118 participants were included in this study, and ILAs were observed with the CT scans of 120 individuals (4%). The median extent of ILA calculated by the quantitative system was 5.8% for the ILA group, 0.7% for the equivocal ILA group, and 0.1% for the no ILA group (P < .001). A 1.8% area threshold in a lung zone for quantitative detection of ILA showed 100% sensitivity and 99% specificity. Progression was observed in 48% of visually assessed fibrotic ILAs (15 of 31), and quantitative extent of ILA increased by 3.1% in subjects with progression. Conclusion ILAs were detected in 4% of the Korean lung cancer screening population. Deep learning-based texture analysis showed high sensitivity and specificity for detecting ILA with use of a 1.8% lung area cutoff value. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Egashira and Nishino in this issue.

Electrical Impedance Tomographic Mapping of Hypoventilated Lung Areas in Intubated Patients With COVID-19.

Electrical Impedance Tomographic Mapping of Hypoventilated Lung Areas in Intubated Patients With COVID-19.

Comparative genomics between matched solid and lepidic portions of semi-solid lung adenocarcinomas

BackgroundGenetic changes that drive the transition from lepidic to invasive cancer development within a radiographic ground glass or semi-solid lung lesion (SSL) are not well understood. Biomarkers to predict the transition to solid, invasive cancer within SSL are needed. MethodsPatients with surgically resected SSL were identified retrospectively from a surgical database. Clinical characteristics and survival were compared between stage I SSL (n = 65) and solid adenocarcinomas (n = 120) resected during the same time period. Areas of normal lung, in situ lepidic, and invasive solid tumor were microdissected from within the same SSL specimens and next generation sequencing (NGS) and Affymetrix microarray of gene expression were performed. ResultsThere were more never smokers, Asian patients, and sub-lobar resections among SSL but no difference in 5-year survival between SSL and solid adenocarcinoma. Driver mutations found in both lepidic and solid invasive portion were EGFR (43%), KRAS (21%), and DNMT3A (5%). CEACAM5 was the most upregulated gene found in solid, invasive portions of SSL. Lepidic and invasive solid areas had many similarities in gene expression, however there were some significant differences with the gene SPP1 being a unique biomarker for the invasive component of a SSL. ConclusionsCommon lung cancer driver mutations are present in in situ lepidic as well as invasive solid portions of a SSL, suggesting early development of driver mutations. CEACAM5 and SPP1 emerged as promising biomarkers of invasive potential in semi-solid lesions. Other studies have shown both genes to correlate with poor prognosis in lung cancer and their role in evolution of semi-solid lung lesions warrants further study.

Enhancing physicians' radiology diagnostics of COVID-19's effects on lung health by leveraging artificial intelligence.

Introduction: This study aimed to develop an individualized artificial intelligence model to help radiologists assess the severity of COVID-19's effects on patients' lung health. Methods: Data was collected from medical records of 1103 patients diagnosed with COVID-19 using RT- qPCR between March and June 2020, in Hospital Madrid-Group (HM-Group, Spain). By using Convolutional Neural Networks, we determine the effects of COVID-19 in terms of lung area, opacities, and pulmonary air density. We then combine these variables with age and sex in a regression model to assess the severity of these conditions with respect to fatality risk (death or ICU). Results: Our model can predict high effect with an AUC of 0.736. Finally, we compare the performance of the model with respect to six physicians' diagnosis, and test for improvements on physicians' performance when using the prediction algorithm. Discussion: We find that the algorithm outperforms physicians (39.5% less error), and thus, physicians can significantly benefit from the information provided by the algorithm by reducing error by almost 30%.

Automatic measurement of cardiothoracic ratio in chest x-ray images with ProGAN-generated dataset

PurposeThe cardiomegaly can be determined by the cardiothoracic ratio (CTR) which can be measured in a chest x-ray image. It is calculated based on a relationship between a size of heart and a transverse dimension of chest. The cardiomegaly is identified when the ratio is larger than a cut-off threshold. This paper aims to propose a solution to calculate the ratio for classifying the cardiomegaly in chest x-ray images.Design/methodology/approachThe proposed method begins with constructing lung and heart segmentation models based on U-Net architecture using the publicly available datasets with the groundtruth of heart and lung masks. The ratio is then calculated using the sizes of segmented lung and heart areas. In addition, Progressive Growing of GANs (PGAN) is adopted here for constructing the new dataset containing chest x-ray images of three classes including male normal, female normal and cardiomegaly classes. This dataset is then used for evaluating the proposed solution. Also, the proposed solution is used to evaluate the quality of chest x-ray images generated from PGAN.FindingsIn the experiments, the trained models are applied to segment regions of heart and lung in chest x-ray images on the self-collected dataset. The calculated CTR values are compared with the values that are manually measured by human experts. The average error is 3.08%. Then, the models are also applied to segment regions of heart and lung for the CTR calculation, on the dataset computed by PGAN. Then, the cardiomegaly is determined using various attempts of different cut-off threshold values. With the standard cut-off at 0.50, the proposed method achieves 94.61% accuracy, 88.31% sensitivity and 94.20% specificity.Originality/valueThe proposed solution is demonstrated to be robust across unseen datasets for the segmentation, CTR calculation and cardiomegaly classification, including the dataset generated from PGAN. The cut-off value can be adjusted to be lower than 0.50 for increasing the sensitivity. For example, the sensitivity of 97.04% can be achieved at the cut-off of 0.45. However, the specificity is decreased from 94.20% to 79.78%.

Changes in ventilation distribution during general anesthesia measured with EIT in mechanically ventilated small children

BackgroundAtelectasis during general anesthesia is a risk for perioperative complications. EIT measurements were performed in mechanically ventilated healthy children during elective surgery to demonstrate the changes in ventilation distribution during general anesthesia. The ventilation distribution was quantified by calculating the Global Inhomogeneity index (GI).MethodsEIT measurements were performed in 23 children (9 weeks—10 years) without lung disease to detect changes in regional ventilation during elective surgery. Three previously defined time points were marked during the measurement: after intubation and start of pressure-controlled ventilation (PCV), change to pressure support ventilation (PSV), and after extubation (spontaneous breathing—SB). Ventilation distribution based on regions of interest (ROI) and changes in end-expiratory volume (∆EELV) were collected at these time points and compared. The Global Inhomogeneity index was calculated at the beginning of pressure-controlled ventilation (PCV).ResultsWith increasing spontaneous breathing, dorsal recruitment of atelectasis occurred. The dorsal ventilation fraction increased over the time of general anesthesia with increasing spontaneous breathing, whereas the ventral fraction decreased relatively (Difference ± 5.5 percentage points respectively; 95% CI; 3.5—7.4; p < 0.001). With the onset of spontaneous breathing, there was a significant reduction in end-expiratory volume (Difference: 105 ml; 95% CI, 75–135; p < 0.001). The GI of the lung-healthy ventilated children is 47% (SD ± 4%).ConclusionControlled ventilation of healthy children resulted in increased ventilation of the ventral and collapse of the dorsal lung areas. Restart of spontaneous breathing after cessation of surgery resulted in an increase in ventilation in the dorsal with decrease in the ventral lung areas. By calculating the GI, representing the ratio of more to less ventilated lung areas, revealed the presumed homogeneous distribution of ventilation.Trial registrationClinicalTrials.gov Registration ID: NCT04873999. First registration: 05/05/2021.

Abstract 5427: Improving non-small cell lung cancer segmentation on a challenging dataset

Abstract When applied to different datasets, performance of the same deep learning tumor segmentation model can greatly vary. In a non-small cell lung cancer CT scan segmentation study that consists of two datasets, we found that the SwinUNETR model achieves state-of-the-art DICE score on a public dataset NSCLC but performs badly on a private dataset of curated data collected clinically. This performance variation reduces the applicability of such models. To mitigate this gap, through experimentation, we identified a set of techniques and applied them in the following order: (1) normalize a dataset to reduce differences between images. (2) stratify a dataset according to tumor sizes to form a more diverse training set. (3) isolate the lung area before training to help the model focus on the right area. (4) before training, initialize models with self-supervised pre-training weights (5) use a new loss function to give more weights on the cancerous area (6) after a model is trained, perform 3-axis test time flipping augmentation and ensemble the final predictions. In our experiments, our set of techniques improved the test DICE score for both datasets we tested on, where the best improvement was a 53% improvement from 0.32 to 0.49 DICE score. Citation Format: Yi Wei, Balaji Selvaraj, Mayank Patwari, Qin Li, Meng Xu, Konstantinos Sidiropoulos, Zhenning Zhang, Leon Fedden, Anant Madabhushi, Mohammadhadi Khorrami, Vidya Sankar Viswanathan, Amit Gupta. Improving non-small cell lung cancer segmentation on a challenging dataset. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5427.